Anomalous Trace Shape Classification in Substrate Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately detect and classify anomalies in substrate processing systems, leading to inefficiencies and inconsistencies in semiconductor device manufacturing.
Innovation Solution
A system that utilizes multiple predictive subsystems, including machine-learning models trained with ensemble techniques and trace analysis methods, to detect anomalies and classify trace shapes, thereby identifying the type of issue causing the anomaly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple predictive subsystems with different techniques are used to detect anomalies, then anomaly detection accuracy is improved, but device complexity increases
Solution Approach 1:
The system divides anomaly detection into multiple independent predictive subsystems, each specializing in different detection techniques. These subsystems process trace data separately and their results are integrated, allowing each component to focus on specific aspects of anomaly detection while maintaining overall system accuracy.
Solution Approach 2:
Multiple predictive subsystems using different detection techniques are combined into a unified system. The outputs from various subsystems are integrated to produce a comprehensive anomaly detection result, leveraging the strengths of each individual technique to achieve higher overall accuracy.
2Reliability
If multiple predictive subsystems are implemented for comprehensive anomaly detection, then reliability is improved, but device complexity increases
Solution Approach 1:
The reliable anomaly detection function is segmented across multiple predictive subsystems, each employing different detection techniques. This distribution of functionality across independent components enhances overall system reliability through diversity while managing complexity through modular architecture.
Solution Approach 2:
Different predictive subsystems utilize varying detection parameters and techniques (e.g., different machine learning algorithms, statistical methods, or threshold criteria). By changing and diversifying these detection parameters across subsystems, the system achieves more reliable anomaly detection through multiple perspectives.
3Manufacturing precision
If automated classification of trace shapes is implemented, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The system implements automated classification of trace shapes using predictive subsystems that independently analyze and categorize anomaly patterns. This self-service classification eliminates manual intervention, improving manufacturing precision through consistent, repeatable classification while the automation manages the complexity of classification logic.
Data Source
AI summary
A system configured to obtain current trace data associated with a substrate processing system and provide the trace data as input to a first predictive subsystem trained to detect anomalies using a first technique. Responsive to detecting an anomaly in the trace data, the system provides the trace data as input to a second predictive subsystem trained to detect anomalies using a second technique. Output data obtained from the first predictive subsystem and the second predictive subsystem is provided to a third predictive subsystem. Output data from the third predictive subsystem is obtained. The output data is reflective of a trace shape associated with the anomaly. Based on the trace shape, a type of issue that caused the anomaly in the trace data is identified.